This paper presents the concept of the project CARE (Safe Private Homes for Elderly Persons) aiming to realise an intelligent monitoring and alarming system for independent living of elderly persons. Specifically, CARE targets the automated recognition and alarming of critical situations (like fall detection) using vision sensors and real-time processing while preserving the privacy and taking into account system dependability issues, especially ensuring reliability, availability, security, and safety from a holistic point of view. A biologically-inspired dynamic stereo vision sensor from AIT will be integrated in the Everon (alarm, security, and monitoring) system for the seamless analysis and tracking of elderly persons’ behaviour at home. This real-time information can be exploited for incident detection (e.g., fall detection, immobilised person), and instantaneous alarming the concerned parties. The system will be tested in pilot installations in Germany at the Residenz Gruppe Bremen and in Finland at Yrjo & Hanna foundation. The project status and early results are presented in this paper.